This replication and expansion of a pilot study focuses upon the prediction of the relative frequency of AWOL events with respect to a cohort of approximately 400 boys placed by the court at an open residential treatment facility in upstate New York. The replication involves the comparison of both psychological information (obtained from boys on a self-report basis) and social background data in a multivariate regression design which will permit an assessment of the relative importance of the two types of information. The expansion of the study involves the use of Q factor analysis techniques in order to divide the study population into a limited number of empirical types. Type membership will then be used as a predictor of AWOL rate. A final expansion of the study involves the exploration of living unit context as a possible determinant of AWOL behavior.